torch.load issue on different folder

[Zhaoming]

I trained 2 set of model and place them on two folder.

One of them can not be loaded when I call them by a program runs on another folder, even though I can detect them all is exist.

but both of them could be loaded when the program runs on their parent folder? why?